While reset and checkout have different usages most of the time, I can't see what difference there is between these two.
There probably is one or nobody would have bothered adding a --hard option to do something the basic checkout can do.
Maybe there is a difference is the way you will see the history?
This answer is mostly quoted from my answer to a previous question: git reset in plain english.
The two are very different. They result in the same state for your index and work tree, but the resulting history and current branch aren't the same.
Suppose your history looks like this, with the master branch currently checked out:
- A - B - C (HEAD, master)
and you run git reset --hard B. You'll get this:
- A - B (HEAD, master) # - C is still here, but there's no
# branch pointing to it anymore
You'd actually get that effect if you use --mixed or --soft too - the only difference is what happens to your work tree and index. In the --hard case, the work tree and index match B.
Now, suppose you'd run git checkout B instead. You'd get this:
- A - B (HEAD) - C (master)
You've ended up in a detached HEAD state. HEAD, work tree, index all match B, same as with the hard reset, but the master branch was left behind at C. If you make a new commit D at this point, you'll get this, which is probably not what you want:
- A - B - C (master)
\
D (HEAD)
So, you use checkout to, well, check out that commit. You can fiddle with it, do what you like, but you've left your branch behind. If you want the branch moved too, you use reset.
If documentation provided with Git doesn't help you, take a look at A Visual Git Reference by Mark Lodato.
In particular if you are comparing git checkout <non-branch> with git reset --hard <non-branch> (hotlinked):
(source: github.com)
Note that in the case of git reset --hard master~3 you leave behind a part of DAG of revisions - some of commits are not referenced by any branch. Those are protected for (by default) 30 days by reflog; they would ultimately be pruned (removed).
git-reset hash sets the branch reference to the given hash, and optionally checks it out, with--hard.
git-checkout hash sets the working tree to the given hash; and unless hash is a branch name, you'll end up with a detached head.
ultimately, git deals with 3 things:
working tree (your code)
-------------------------------------------------------------------------
index/staging-area
-------------------------------------------------------------------------
repository (bunch of commits, trees, branch names, etc)
git-checkout by default just updates the index and the working tree, and can optionally update something in the repository (with the -b option)
git-reset by default just updates the repository and the index, and optionally the working tree (with the --hard option)
You can think of the repository like this:
HEAD -> master
refs:
master -> sha_of_commit_X
dev -> sha_of_commit_Y
objects: (addressed by sha1)
sha_of_commit_X, sha_of_commit_Y, sha_of_commit_Z, sha_of_commit_A ....
git-reset manipulates what the branch references point to.
Suppose your history looks like this:
T--S--R--Q [master][dev]
/
A--B--C--D--E--F--G [topic1]
\
Z--Y--X--W [topic2][topic3]
Keep in mind that branches are just names that advance automatically when you commit.
So you have the following branches:
master -> Q
dev -> Q
topic1 -> G
topic2 -> W
topic3 -> W
And your current branch is topic2, that is, the HEAD points to topic2.
HEAD -> topic2
Then, git reset X will reset the name topic2 to point to X; meaning if you make a commit P on branch topic2, things will look like this:
T--S--R--Q [master][dev]
/
A--B--C--D--E--F--G [topic1]
\
Z--Y--X--W [topic3]
\
P [topic2]
